About this Document

Using the html file

On the left is a floating table of contents for navigating between major sections of output:

  • Exploration and QA/QC
  • Summary bar charts
  • Univariate analyses
  • Multivariate analyses

Within each major section, content is split into tabs that are spread horizontally across the screen. By default, the About tab for each section is showing. Click on any tab to see specifics.

General notes about notation, data inclusion, and reserve choices

  • If a species was identified as being of interest, and the category that species belongs to (e.g. H-Halophyte / B-Brackish / Unvegetated category) was also identified as being of interest, the species takes precedent and the category represents only the other species in that category. The species itself is not represented in both.
  • If a species was identified as belonging to an “Other layer” - something like canopy wrack, or water when whatever is below the water is also measured - it was removed before any of these analyses (including exploratory graphics).

Exploration and QA/QC

About

This section contains tables and graphs, in separate tabs.

To detect errors in data sets, the information in this section should be reviewed by Reserve data practitioners. These results are meant to help you find issues that need to be corrected in the raw data, and decide whether plots or data should be removed from analyses (e.g., “off season” surveys, plots no longer monitored, restoration sites).

Ask yourself:

  • Do you see the correct number of sampling events per year? If not, what is missing or extra? This implies a correction needs to be made in the raw data file.
  • Do you see the correct number of plots per site?
  • Are species names spelled correctly? Are there any close variants or duplicates (e.g., capital vs lowercase, misspelled)?
  • General percent cover in plots - is anything unusual (e.g., over 100% when standardized, sudden drops or spikes in species)?

Missing/Removed Data

Plots/dates without samples

There were 0 rows with no data. These rows were removed from the dataset before further processing. If rows were removed, relevant information is in a table below.

Plots without enough samples

If monitoring plots did not have data in at least 3 separate years, they were removed from the dataset before statistical analyses. These plots do appear in the plots and information in the Exploration/QAQC and Summary Bar Chart sections of this document, but were removed before Univariate and Multivariate analyses.

In this dataset, there were 0 such plots removed. If plots were removed, relevant information is in a table below.

Data flagged suspect or reject

There were 0 data points that were removed due to QA/QC flags. If data points were removed, relevant information is in a table below.

Sampling Info

Samples per year

# samples per year, by site
Site 2015 2016 2017 2018 2019 2020
clmaj 0 30 30 30 30 30
juro high 21 21 21 23 21 21
juro low 35 35 35 35 35 35
juro mid 40 40 40 40 40 40
spalt 27 27 27 27 27 27
# samples per year, by vegetation zone
Vegetation_Zone 2015 2016 2017 2018 2019 2020
L-Low Marsh 22 22 22 22 22 22
T-Transition 87 87 87 89 87 87
P-Pools/Pannes 5 5 5 5 5 5
H-High Marsh 6 6 6 6 6 6
UE-Upland Edge 3 3 3 3 3 3
FT-Freshwater Tidal 0 30 30 30 30 30

Species

Plots and Zones

Make sure the colors below correctly represent which Vegetation Zones your plots belong to. If something looks wrong, it needs to be corrected in the ‘Station_Table’ tab of your data workbook.

Time-series - Species

By Zone

All plots combined

EMI

EMI, Ecotone Migration Index, is the proportional cover of species/covers or species/cover groupings that are expected to increase within a vegetation zone as sea level rises. These species were identified by reserve staff, for each zone, in the ‘veg-specs.xlsx’ file.

Species considered to be ‘migrators’ within each zone are denoted by ‘x’ in the following table:

Species L-Low Marsh T-Transition P-Pools/Pannes H-High Marsh UE-Upland Edge
Unvegetated X X X X
Spartina alterniflora X X X
Juncus roemerianus X X X
Agalinus maritima X
Baccharis halimifolia X
Batis maritima X
Borrichia frutescens X
Distichlis spicata X
Ipomoea sagittata X
Ipomoea sp. X
Iva frutescens X
Juncus sp. X
Limonium carolinianum X
Sabatia stellaris X
Salicornia depressa X
Spartina patens X
Spartina spartinae X
Sphagnum sp. X

Summary Bar Charts

About

This section contains summary graphics. The color palettes are generated via the khroma R package and were developed to be colorblind-friendly.

These summary figures tell a graphical story about current conditions and trends through time. There are two types of figure:

  • Averaged stacked bar charts - show the relative distribution of species and cover classes and how these relationships change over time.
    • Charts have been created at the Site, Zone, and Site x Zone levels.
    • Chart categories feature dominant species or species groups as identified by the Reserve in the “veg-specs.xlsx” file, “Analysis_Specs” sheet.
  • Spatial stacked bar charts - show the relative distribution of species and cover classes as above, but for each plot. These charts are laid out spatially by site. For ease of interpretation, only 4 evenly-spaced time points are used along the x-axis in each chart.

Questions addressed

  • Which species characterize each site and marsh zone?
  • How do relative abundances of species/groups fluctuate among years and is there a visual trend with time (e.g., corresponds to severe events, storms, staff turnover)?
  • For QA/QC purposes, does the data make sense?

Averaged: By Veg Zone

User-chosen species

The species/groups identified by the Reserve for these graphs are (in order):

  • Spartina alterniflora
  • Juncus roemerianus
  • Salicornia depressa
  • Cladium jamaicense

And so the groups appearing in the plots should be (alphabetically):

  • Cladium jamaicense
  • Juncus roemerianus
  • Salicornia depressa
  • Spartina alterniflora
  • Other

Top 3 species by mean cover across the entire dataset

These were automatically calculated as the species having the highest mean cover, across all plots and dates.

Averaged: By Site

User-chosen species

Top 3 species by mean cover across the entire dataset

Averaged: Zone within Site

User-chosen species

Top 3 species by mean cover across the entire dataset

Spatial

One spatial graph was generated for each Site. This can be changed between the options of Site and Transect in the “More_Options” sheet of the “veg-specs.xlsx” file.

Each panel on the graphs represents an individual vegetation plot, showing 4 evenly spaced-through-time samples, of the species choices from the “Analysis_Specs” sheet of the “veg-specs.xlsx” file.

Plots without associated spatial locations will not be included in these graphics.

User-chosen species

Top 3 species by mean cover across the entire dataset

Univariate Analyses

About

A tab will be generated for results of each response variable. The response variables consist of each of the (up to) 4 response variables identified in the Analysis Specs spreadsheet, (up to) 2 custom metrics identified in the Analysis Specs spreadsheet, EMI (Ecotone Migration Index), Species Richness, and Shannon-Weiner Diversity Index. You will see the same tables and graphics for each.

Questions addressed

  • Are there shifts in key vegetation species/groups over time?
  • Do these shifts vary by vegetation zone?
  • Follow-up: Where is the plant community changing (which plots) and what characteristics do those areas have in common (e.g. site, zone, distance from water, elevation)?

Statistical Model

The main statistical model used is a linear mixed model, via lme4::lmer(), with the form y ~ Vegetation Zone + Time + Zone*Time + 1|Plot, where y is the response variable given in the tab, and each individual plot has a random intercept.

If only one vegetation zone is present in the data file, Vegetation Zone is removed and the statistical model is simplified to y ~ Time + 1|Plot.

Reserve-identified response variables:

The reserve-identified univariate response variables are:

  • Spartina alterniflora
  • Juncus roemerianus
  • Salicornia depressa
  • Cladium jamaicense

Custom Metrics

Any custom metrics specified in the ‘Analysis_Specs’ sheet of the ‘veg-specs.xlsx’ file are calculated here, for inclusion in univariate analyses.

If you do not see a tab for a metric you believe you defined, look just below these bullet points to see if there are error messages showing up in little boxes that are different from other text - these may be helpful in troubleshooting. There are a few likely reasons for problems:

  • Make sure the species or groupings you want to use are in the ‘Species_Names’ sheet of your data file.
  • Make sure any species or groupings are enclosed in backticks. Backticks are not the same as single quotes or apostrophes. To make a backtick, use the key to the left of the number 1 on your keyboard.
  • Make sure species and groups are spelled correctly.
  • Make sure species and groups are capitalized correctly (“correctly” = however it is capitalized in the ‘Species_Names’ sheet of the data file). e.g., if you want to include ‘live vegetation’ in your metric, check the Species_Names sheet of the data file and you will see it is ‘Live vegetation’. If you capitalize it any other way, R will not recognize it and thus will not calculate the metric.
  • Check your math - if you are using division, try adding 1 to the denominator to avoid the possibility of dividing by 0. Make sure you have parentheses around terms that should be evaluated together.

EMI

EMI, Ecotone Migration Index, is the proportional cover of species/covers or species/cover groupings that are expected to increase within a vegetation zone as sea level rises. These species were identified by reserve staff, for each zone, in the ‘veg-specs.xlsx’ file.

See the ‘EMI’ tab in the ‘Exploratory and QA/QC’ section of this document for a table of species considered to be ‘ecotone migrators’ for each zone.

The current file contains multiple vegetation zones, and vegetation zone will be one of the predictors in the statistical model. Results are presented for each vegetation zone independently because the marginal effects are of interest to the project team. However, be careful interpreting zone-wise results if the interaction term (Years_sinceStart:Vegetation_Zone) is not significant (p > 0.05).

Numerical effects

EMI - Type III Analysis of Variance Table, using Kenward-Rogers method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Years_sinceStart 0.4475 0.4475 1 731.16 18.13 0
Vegetation_Zone 4.7897 0.9579 5 475.24 38.80 0
Years_sinceStart:Vegetation_Zone 1.9051 0.3810 5 731.16 15.43 0
EMI - Estimated marginal slopes (change per year for each zone)
Veg Zone Trend (per year) SE CI lower CI upper t df p.val
L-Low Marsh 0.0485 0.0081 0.033 0.064 6.0 731.1 0.0000
T-Transition 0.0725 0.0041 0.064 0.081 17.7 731.2 0.0000
P-Pools/Pannes -0.0053 0.0173 -0.039 0.029 -0.3 731.2 0.7596
H-High Marsh 0.0541 0.0157 0.023 0.085 3.4 731.2 0.0006
UE-Upland Edge -0.0216 0.0223 -0.065 0.022 -1.0 731.1 0.3341
FT-Freshwater Tidal 0.0000 0.0094 -0.018 0.018 0.0 731.1 1.0000
Marginal R^2 (R2m) represents the variance explained by the fixed effects. Conditional R^2 (R2c) represents the variance explained by the entire model, which includes both fixed and random effects.
EMI - Margn’l and Cond’l R^2 values
R2m R2c
0.664 0.738

Graphs

Contrasts plot

If the interaction between vegetation zone and time was significant (p < 0.05), letters are used to represent groups of slopes that are not significantly different (via pairwise comparisons). If the interaction was not significant, this plot represents the marginal slopes but no letters are used.

Some notes from the function. Kenward-Roger method used to estimate denominator degrees of freedom. Confidence level used: 0.95. p-value adjustment: Tukey. alpha = 0.05.

“If two or more means share the same grouping symbol, then we cannot show them to be different. But we also did not show them to be the same.”

Same plot, but x-axis goes from lowest to highest zone.

Richness

Species Richness, calculated on all non-abiotic, non-dead, and non-overstory columns; using the vegan package.

The current file contains multiple vegetation zones, and vegetation zone will be one of the predictors in the statistical model. Results are presented for each vegetation zone independently because the marginal effects are of interest to the project team. However, be careful interpreting zone-wise results if the interaction term (Years_sinceStart:Vegetation_Zone) is not significant (p > 0.05).

Numerical effects

Richness - Type III Analysis of Variance Table, using Kenward-Rogers method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Years_sinceStart 160.0977 160.0977 1 731.06 123.45 0.0000
Vegetation_Zone 18.8745 3.7749 5 325.96 2.91 0.0138
Years_sinceStart:Vegetation_Zone 176.1839 35.2368 5 731.06 27.17 0.0000
Richness - Estimated marginal slopes (change per year for each zone)
Veg Zone Trend (per year) SE CI lower CI upper t df p.val
L-Low Marsh 0.0368 0.0586 -0.078 0.152 0.6 731.0 0.5304
T-Transition 0.1199 0.0298 0.061 0.178 4.0 731.1 0.0001
P-Pools/Pannes 0.3238 0.1250 0.078 0.569 2.6 731.1 0.0098
H-High Marsh 0.6881 0.1139 0.465 0.912 6.0 731.1 0.0000
UE-Upland Edge 0.7870 0.1618 0.469 1.105 4.9 731.0 0.0000
FT-Freshwater Tidal 0.8493 0.0680 0.716 0.983 12.5 731.0 0.0000
Marginal R^2 (R2m) represents the variance explained by the fixed effects. Conditional R^2 (R2c) represents the variance explained by the entire model, which includes both fixed and random effects.
Richness - Margn’l and Cond’l R^2 values
R2m R2c
0.446 0.666

Graphs

Contrasts plot

If the interaction between vegetation zone and time was significant (p < 0.05), letters are used to represent groups of slopes that are not significantly different (via pairwise comparisons). If the interaction was not significant, this plot represents the marginal slopes but no letters are used.

Some notes from the function. Kenward-Roger method used to estimate denominator degrees of freedom. Confidence level used: 0.95. p-value adjustment: Tukey. alpha = 0.05.

“If two or more means share the same grouping symbol, then we cannot show them to be different. But we also did not show them to be the same.”

Same plot, but x-axis goes from lowest to highest zone.

Diversity

Shannon-Weiner Diversity index, calculated on all non-abiotic, non-dead, and non-overstory columns; using the vegan package.

The current file contains multiple vegetation zones, and vegetation zone will be one of the predictors in the statistical model. Results are presented for each vegetation zone independently because the marginal effects are of interest to the project team. However, be careful interpreting zone-wise results if the interaction term (Years_sinceStart:Vegetation_Zone) is not significant (p > 0.05).

Numerical effects

SWdiv - Type III Analysis of Variance Table, using Kenward-Rogers method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Years_sinceStart 12.2815 12.2815 1 731.03 146.03 0.0000
Vegetation_Zone 0.9379 0.1876 5 269.64 2.23 0.0516
Years_sinceStart:Vegetation_Zone 11.6558 2.3312 5 731.03 27.72 0.0000
SWdiv - Estimated marginal slopes (change per year for each zone)
Veg Zone Trend (per year) SE CI lower CI upper t df p.val
L-Low Marsh 0.0297 0.0149 0.000 0.059 2.0 731.0 0.0469
T-Transition 0.0419 0.0076 0.027 0.057 5.5 731.0 0.0000
P-Pools/Pannes 0.0624 0.0318 0.000 0.125 2.0 731.1 0.0505
H-High Marsh 0.1918 0.0290 0.135 0.249 6.6 731.0 0.0000
UE-Upland Edge 0.2196 0.0412 0.139 0.300 5.3 731.0 0.0000
FT-Freshwater Tidal 0.2315 0.0173 0.198 0.266 13.4 731.0 0.0000
Marginal R^2 (R2m) represents the variance explained by the fixed effects. Conditional R^2 (R2c) represents the variance explained by the entire model, which includes both fixed and random effects.
SWdiv - Margn’l and Cond’l R^2 values
R2m R2c
0.412 0.709

Graphs

Contrasts plot

If the interaction between vegetation zone and time was significant (p < 0.05), letters are used to represent groups of slopes that are not significantly different (via pairwise comparisons). If the interaction was not significant, this plot represents the marginal slopes but no letters are used.

Some notes from the function. Kenward-Roger method used to estimate denominator degrees of freedom. Confidence level used: 0.95. p-value adjustment: Tukey. alpha = 0.05.

“If two or more means share the same grouping symbol, then we cannot show them to be different. But we also did not show them to be the same.”

Same plot, but x-axis goes from lowest to highest zone.

Spartina alterniflora

The current file contains multiple vegetation zones, and vegetation zone will be one of the predictors in the statistical model. Results are presented for each vegetation zone independently because the marginal effects are of interest to the project team. However, be careful interpreting zone-wise results if the interaction term (Years_sinceStart:Vegetation_Zone) is not significant (p > 0.05).

Numerical effects

Spartina alterniflora - Type III Analysis of Variance Table, using Kenward-Rogers method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Years_sinceStart 150.2792 150.2792 1 731.03 3.71 0.0545
Vegetation_Zone 12186.0749 2437.2150 5 260.03 60.15 0.0000
Years_sinceStart:Vegetation_Zone 1918.1691 383.6338 5 731.03 9.47 0.0000
Spartina alterniflora - Estimated marginal slopes (change per year for each zone)
Veg Zone Trend (per year) SE CI lower CI upper t df p.val
L-Low Marsh -2.5657 0.3274 -3.208 -1.923 -7.8 731 0.0000
T-Transition -0.1517 0.1665 -0.479 0.175 -0.9 731 0.3624
P-Pools/Pannes 0.0000 0.6988 -1.372 1.372 0.0 731 1.0000
H-High Marsh 0.0000 0.6364 -1.249 1.249 0.0 731 1.0000
UE-Upland Edge 0.0000 0.9044 -1.776 1.776 0.0 731 1.0000
FT-Freshwater Tidal 0.0000 0.3802 -0.746 0.746 0.0 731 1.0000
Marginal R^2 (R2m) represents the variance explained by the fixed effects. Conditional R^2 (R2c) represents the variance explained by the entire model, which includes both fixed and random effects.
Spartina alterniflora - Margn’l and Cond’l R^2 values
R2m R2c
0.52 0.774

Graphs

Contrasts plot

If the interaction between vegetation zone and time was significant (p < 0.05), letters are used to represent groups of slopes that are not significantly different (via pairwise comparisons). If the interaction was not significant, this plot represents the marginal slopes but no letters are used.

Some notes from the function. Kenward-Roger method used to estimate denominator degrees of freedom. Confidence level used: 0.95. p-value adjustment: Tukey. alpha = 0.05.

“If two or more means share the same grouping symbol, then we cannot show them to be different. But we also did not show them to be the same.”

Same plot, but x-axis goes from lowest to highest zone.

Juncus roemerianus

The current file contains multiple vegetation zones, and vegetation zone will be one of the predictors in the statistical model. Results are presented for each vegetation zone independently because the marginal effects are of interest to the project team. However, be careful interpreting zone-wise results if the interaction term (Years_sinceStart:Vegetation_Zone) is not significant (p > 0.05).

Numerical effects

Juncus roemerianus - Type III Analysis of Variance Table, using Kenward-Rogers method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Years_sinceStart 1954.054 1954.054 1 731.08 8.20 0.0043
Vegetation_Zone 82119.867 16423.973 5 373.70 68.92 0.0000
Years_sinceStart:Vegetation_Zone 24887.239 4977.448 5 731.08 20.89 0.0000
Juncus roemerianus - Estimated marginal slopes (change per year for each zone)
Veg Zone Trend (per year) SE CI lower CI upper t df p.val
L-Low Marsh -2.3022 0.7940 -3.861 -0.743 -2.9 731.1 0.0039
T-Transition -7.4206 0.4038 -8.213 -6.628 -18.4 731.1 0.0000
P-Pools/Pannes -0.5468 1.6947 -3.874 2.780 -0.3 731.1 0.7471
H-High Marsh 0.0955 1.5433 -2.934 3.125 0.1 731.1 0.9507
UE-Upland Edge 0.0000 2.1935 -4.306 4.306 0.0 731.1 1.0000
FT-Freshwater Tidal 0.3752 0.9221 -1.435 2.186 0.4 731.1 0.6842
Marginal R^2 (R2m) represents the variance explained by the fixed effects. Conditional R^2 (R2c) represents the variance explained by the entire model, which includes both fixed and random effects.
Juncus roemerianus - Margn’l and Cond’l R^2 values
R2m R2c
0.509 0.67

Graphs

Contrasts plot

If the interaction between vegetation zone and time was significant (p < 0.05), letters are used to represent groups of slopes that are not significantly different (via pairwise comparisons). If the interaction was not significant, this plot represents the marginal slopes but no letters are used.

Some notes from the function. Kenward-Roger method used to estimate denominator degrees of freedom. Confidence level used: 0.95. p-value adjustment: Tukey. alpha = 0.05.

“If two or more means share the same grouping symbol, then we cannot show them to be different. But we also did not show them to be the same.”

Same plot, but x-axis goes from lowest to highest zone.

Salicornia depressa

The current file contains multiple vegetation zones, and vegetation zone will be one of the predictors in the statistical model. Results are presented for each vegetation zone independently because the marginal effects are of interest to the project team. However, be careful interpreting zone-wise results if the interaction term (Years_sinceStart:Vegetation_Zone) is not significant (p > 0.05).

Numerical effects

Salicornia depressa - Type III Analysis of Variance Table, using Kenward-Rogers method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Years_sinceStart 12.4416 12.4416 1 731.03 60.41 0.0000
Vegetation_Zone 0.9854 0.1971 5 266.86 0.96 0.4448
Years_sinceStart:Vegetation_Zone 53.6092 10.7218 5 731.03 52.06 0.0000
Salicornia depressa - Estimated marginal slopes (change per year for each zone)
Veg Zone Trend (per year) SE CI lower CI upper t df p.val
L-Low Marsh 0.0000 0.0233 -0.046 0.046 0.0 731 1.0000
T-Transition -0.0211 0.0119 -0.044 0.002 -1.8 731 0.0755
P-Pools/Pannes 0.8030 0.0498 0.705 0.901 16.1 731 0.0000
H-High Marsh 0.0000 0.0454 -0.089 0.089 0.0 731 1.0000
UE-Upland Edge 0.0000 0.0645 -0.127 0.127 0.0 731 1.0000
FT-Freshwater Tidal 0.0000 0.0271 -0.053 0.053 0.0 731 1.0000
Marginal R^2 (R2m) represents the variance explained by the fixed effects. Conditional R^2 (R2c) represents the variance explained by the entire model, which includes both fixed and random effects.
Salicornia depressa - Margn’l and Cond’l R^2 values
R2m R2c
0.257 0.638

Graphs

Contrasts plot

If the interaction between vegetation zone and time was significant (p < 0.05), letters are used to represent groups of slopes that are not significantly different (via pairwise comparisons). If the interaction was not significant, this plot represents the marginal slopes but no letters are used.

Some notes from the function. Kenward-Roger method used to estimate denominator degrees of freedom. Confidence level used: 0.95. p-value adjustment: Tukey. alpha = 0.05.

“If two or more means share the same grouping symbol, then we cannot show them to be different. But we also did not show them to be the same.”

Same plot, but x-axis goes from lowest to highest zone.

Cladium jamaicense

The current file contains multiple vegetation zones, and vegetation zone will be one of the predictors in the statistical model. Results are presented for each vegetation zone independently because the marginal effects are of interest to the project team. However, be careful interpreting zone-wise results if the interaction term (Years_sinceStart:Vegetation_Zone) is not significant (p > 0.05).

Numerical effects

Cladium jamaicense - Type III Analysis of Variance Table, using Kenward-Rogers method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Years_sinceStart 2945.338 2945.338 1 731.07 52.04 0
Vegetation_Zone 21108.025 4221.605 5 355.60 74.59 0
Years_sinceStart:Vegetation_Zone 6325.105 1265.021 5 731.07 22.35 0
Cladium jamaicense - Estimated marginal slopes (change per year for each zone)
Veg Zone Trend (per year) SE CI lower CI upper t df p.val
L-Low Marsh 0.0000 0.3870 -0.760 0.760 0.0 731.1 1.0000
T-Transition -0.0203 0.1968 -0.407 0.366 -0.1 731.1 0.9180
P-Pools/Pannes 0.0000 0.8259 -1.621 1.621 0.0 731.1 1.0000
H-High Marsh -0.5433 0.7521 -2.020 0.933 -0.7 731.1 0.4703
UE-Upland Edge -7.3813 1.0690 -9.480 -5.283 -6.9 731.1 0.0000
FT-Freshwater Tidal -4.0854 0.4494 -4.968 -3.203 -9.1 731.1 0.0000
Marginal R^2 (R2m) represents the variance explained by the fixed effects. Conditional R^2 (R2c) represents the variance explained by the entire model, which includes both fixed and random effects.
Cladium jamaicense - Margn’l and Cond’l R^2 values
R2m R2c
0.478 0.662

Graphs

Contrasts plot

If the interaction between vegetation zone and time was significant (p < 0.05), letters are used to represent groups of slopes that are not significantly different (via pairwise comparisons). If the interaction was not significant, this plot represents the marginal slopes but no letters are used.

Some notes from the function. Kenward-Roger method used to estimate denominator degrees of freedom. Confidence level used: 0.95. p-value adjustment: Tukey. alpha = 0.05.

“If two or more means share the same grouping symbol, then we cannot show them to be different. But we also did not show them to be the same.”

Same plot, but x-axis goes from lowest to highest zone.

JuroSpal Ratio

This custom metric was calculated using the following formula:

Juncus roemerianus / (Spartina alterniflora + 1)

The current file contains multiple vegetation zones, and vegetation zone will be one of the predictors in the statistical model. Results are presented for each vegetation zone independently because the marginal effects are of interest to the project team. However, be careful interpreting zone-wise results if the interaction term (Years_sinceStart:Vegetation_Zone) is not significant (p > 0.05).

Numerical effects

JuroSpal Ratio - Type III Analysis of Variance Table, using Kenward-Rogers method
Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
Years_sinceStart 1688.119 1688.119 1 731.08 6.97 0.0085
Vegetation_Zone 81401.697 16280.339 5 365.95 67.22 0.0000
Years_sinceStart:Vegetation_Zone 27734.970 5546.994 5 731.08 22.90 0.0000
JuroSpal Ratio - Estimated marginal slopes (change per year for each zone)
Veg Zone Trend (per year) SE CI lower CI upper t df p.val
L-Low Marsh -1.4817 0.8005 -3.053 0.090 -1.9 731.1 0.0646
T-Transition -7.5498 0.4071 -8.349 -6.751 -18.5 731.1 0.0000
P-Pools/Pannes -0.5470 1.7085 -3.901 2.807 -0.3 731.1 0.7489
H-High Marsh 0.0955 1.5559 -2.959 3.150 0.1 731.1 0.9510
UE-Upland Edge 0.0000 2.2113 -4.341 4.341 0.0 731.1 1.0000
FT-Freshwater Tidal 0.3752 0.9296 -1.450 2.200 0.4 731.1 0.6866
Marginal R^2 (R2m) represents the variance explained by the fixed effects. Conditional R^2 (R2c) represents the variance explained by the entire model, which includes both fixed and random effects.
JuroSpal Ratio - Margn’l and Cond’l R^2 values
R2m R2c
0.491 0.663

Graphs

Contrasts plot

If the interaction between vegetation zone and time was significant (p < 0.05), letters are used to represent groups of slopes that are not significantly different (via pairwise comparisons). If the interaction was not significant, this plot represents the marginal slopes but no letters are used.

Some notes from the function. Kenward-Roger method used to estimate denominator degrees of freedom. Confidence level used: 0.95. p-value adjustment: Tukey. alpha = 0.05.

“If two or more means share the same grouping symbol, then we cannot show them to be different. But we also did not show them to be the same.”

Same plot, but x-axis goes from lowest to highest zone.

Multivariate Analyses

About

These tabs are the multivariate analyses.

In this section, we use multivariate techniques to ordination to visualize and analyze plant community change through plot-level cover data across marsh zones.

Questions and statistical methods

  • Are there shifts in the vegetation community (as defined by the entire percent-cover matrix) over time? Do these shifts vary by vegetation zone?
    • addressed with PERMANOVA, comparing first year of monitoring to the ‘last’ (most recent) year of monitoring.
    • the first and last year of monitoring at a specific vegetation plot should be considered together. To do this, permutations were restricted to allow swapping of time point within a vegetation plot, but keep both time points of a plot together as plots are permuted across vegetation zones. Unfortunately these restricted permutations do not account for the repeated measures within a plot. Ideally we could use a random effect, as in the univariate models. At this point such a model for PERMANOVA is not possible in R.
    • if the above test indicated that differences in time between vegetation zones were significant or nearly so (p <= 0.10), the species matrix was split by Vegetation Zone. A PERMANOVA was run for each vegetation zone, again with restricted permutations to allow only permutation of tiem point within vegetation plot.
  • Which species/groups contribute most to these shifts?
    • addressed using SIMPER, comparing first and last year of monitoring data within each zone. SIMPER is used to follow up on vegetation zones where the p-value for the PERMANOVA was <= 0.2.
  • Where is the plant community changing, and what characteristics do those areas have in common (e.g. site, zone, distance from water, elevation)?
    • visualized via NMDS.

Reserve output choices

The following species/groups were identified by the reserve as important loading factors to display on NMDS outputs.

Sometimes these species do not appear on the plots; usually that is because all data points for that species were 0 or very close to it.

Up to 8 species/groups could be identified specifically in outputs of these multivariate analyses. Reserve choices are:

  • Spartina alterniflora
  • Juncus roemerianus
  • Salicornia depressa
  • Cladium jamaicense

Characteristics of multivariate data frame

Years in ‘Start’ and ‘End’ time groups for each Vegetation Zone. Number in parentheses is number of samples.
Vegetation_Zone Start End
L-Low Marsh 2015 (22) 2020 (22)
T-Transition 2015 (87) 2020 (87)
P-Pools/Pannes 2015 (5) 2020 (5)
H-High Marsh 2015 (6) 2020 (6)
UE-Upland Edge 2015 (3) 2020 (3)
FT-Freshwater Tidal 2016 (30) 2020 (30)
Species/Groups included in response matrix
Unvegetated
Agalinus maritima
Andropogon sp.
Andropogon virginicus var. glaucus
Anthriscus sp.
Asclepias lanceolata
Aster sp.
Baccharis halimifolia
Batis maritima
Bidens mitis
Bidens sp.
Borrichia frutescens
Centella erecta
Cladium jamaicense
Cyperus sp.
Daucus sp.
Dichanthelium sp.
Digitalis sp.
Diodia sp.
Distichlis spicata
Eleocharis tuberculosa
Eriocaulon compressum
Fimbristylis sp.
Fuirena breviseta
Heterotheca sp.
Ilex sp.
Ilex vomitoria
Ipomoea sagittata
Ipomoea sp.
Iva frutescens
Juncus roemerianus
Juncus sp.
Limonium carolinianum
Ludwigia sp.
Mikania scandens
Morella cerifera
Osmunda regalis
Panicum virgatum
Paspalum floridanum
Paspalum sp.
Phyla nodiflora
Pinus elliottii
Pluchea foetida
Pluchea sp.
Polygala ramosa
Proserpinaca sp.
Rhexia lutea
Rhexia sp.
Rhynchospora corniculata
Rhynchospora elliottii
Rhynchospora fascicularis
Rhynchospora rarifolia
Rhynchospora sp.
Sabatia stellaris
Sagittaria lancifolia
Sagittaria latifolia
Salicornia depressa
Sapium sebiferum
Schoenplectus sp.
Scleria sp.
Setaria sp.
Setaria parviflora
Seutera angustifolia
Solidago sp.
Spartina alterniflora
Spartina patens
Spartina spartinae
Sphagnum sp.
Toxicodendron radicans
Triadica sebifera
Unknown
Unknown 2
Verbena sp.
Woodwardia virginica
Xyris laxifolia
Xyris sp.

PERMANOVA

Overall

H0: Community change (if any) between start and end is consistent across vegetation zones.

The interaction p-value (Vegetation_Zone:Time_group) is what to look at here:

  • If the interaction is significant, then something different happened through time in at least one vegetation zone. Do not interpret the main effects; proceed to the zone-wise tables below. If the interaction is significant or close to it (p <= 0.10), separate PERMANOVAs will be run for each vegetation zone to determine whether community change occurred in each zone separately.
  • If the interaction is not significant, look at the main effect of Time_group to determine whether, across all zones, the community was different at the end of monitoring than at the beginning.

Permutations have been restricted so that time points are only permuted within a vegetation plot, and both time points for a plot are permuted together across vegetation zones.

If only one vegetation zone is present in the data file, the overall PERMANOVA will be skipped; look in the zone-wise PERMANOVA section below for results.

Overall PERMANOVA results, terms treated sequentially
Df SumOfSqs R2 F Pr(>F)
Time_group 1 3.4191 0.0833 57.1142 0.001
Vegetation_Zone 5 17.7434 0.4324 59.2790 0.001
Time_group:Vegetation_Zone 5 2.2721 0.0554 7.5909 0.001
Residual 294 17.6000 0.4289
Total 305 41.0345 1.0000

Zone-wise

H0: No community difference between ‘start’ and ‘end’ within a vegetation zone.

Permutations have been restricted so time points are only permuted within a vegetation plot.

Summary of pairwise PERMANOVAs testing community difference between first and most recent years of monitoring
Vegetation Zone R2 p.value p.value_Bonferonni
L-Low Marsh 0.1208 0.013 0.078
T-Transition 0.2781 0.001 0.006
P-Pools/Pannes 0.0226 0.702 1.000
H-High Marsh 0.3401 0.024 0.144
UE-Upland Edge 0.3769 0.200 1.000
FT-Freshwater Tidal 0.2260 0.001 0.006

Check for homogeneity of dispersion

H0: No difference in dispersion between groups.

This is important to check because one of the assumptions of PERMANOVA is homogeneity of dispersion. Dispersion is the multivariate equivalent of variance. If this assumption is violated, caution should be used in interpreting PERMANOVA results.

The test used here is PERMDISP, implemented with the betadisper function of the vegan package.

Results of dispersion test, overall PERMANOVA
Df Sum Sq Mean Sq F N.Perm Pr(>F)
Groups 11 0.58 0.05 2.31 999 0.011
Residuals 294 6.67 0.02

If the PERMDISP indicated significant difference in dispersions, you should further investigate the following outputs:

Group Mean Distance
L-Low Marsh; Start 0.274
L-Low Marsh; End 0.173
T-Transition; Start 0.227
T-Transition; End 0.140
P-Pools/Pannes; Start 0.171
P-Pools/Pannes; End 0.148
H-High Marsh; Start 0.172
H-High Marsh; End 0.186
UE-Upland Edge; Start 0.315
UE-Upland Edge; End 0.159
FT-Freshwater Tidal; Start 0.197
FT-Freshwater Tidal; End 0.217

Summary of dispersion test outputs for each zone
Vegetation Zone Start.dispersion End.dispersion NumDf DenDf F.val N.Perm p.val
L-Low Marsh 0.272 0.175 1 42 4.1 999 0.0570
T-Transition 0.227 0.141 1 172 14.5 999 0.0010
P-Pools/Pannes 0.172 0.153 1 8 0.0 999 0.8310
H-High Marsh 0.177 0.187 1 10 0.0 999 0.8750
UE-Upland Edge 0.320 0.161 1 4 0.7 719 0.6014
FT-Freshwater Tidal 0.198 0.218 1 58 0.3 999 0.6010
Click to view plots of dispersion for each zone


SIMPER

SIMPER was run if:

  • Across all zones: the interaction term in the overall PERMANOVA was not significant (p > 0.05) and the main effect for time was significant or close to it (p <= 0.2).
  • Zone-wise: the interaction term in the overall PERMANOVA was significant or close to it (p <= 0.1) and the within-zone effect for time in the zone-wise PERMANOVA was significant or close to it (p <= 0.2).
  • Generally only one version of the SIMPER will be run (across all zones vs. zone-wise), but when the interaction term was near significance (0.05 < p < 0.1), SIMPER was run both ways.

The p-values determining the above logic are unadjusted. Due to the exploratory nature of these analyses, we did not adjust p-values for multiple comparisons

SIMPER output explanation:

The top 6 species in output are below.

“average” is the average contribution of that species to the Bray-Curtis distance between the two groups (note, this is not expressed in % and the column does not total to 1); “sd” is the standard deviation of the species’ contribution. “cumulative” is the cumulative % contribution for this species and all those above it in the table. Typically people only report species up to the one that brings “cumulative” over 0.7. “p” is a p-value for that species based on permutation tests. “mean_start” is the mean cover of that species in the starting year(s), and “mean_end” is the mean cover of the species in the last year(s) of monitoring.

Results

L-Low Marsh SIMPER results; PERMANOVA p = 0.013
average sd cumulative p mean_start mean_end
Spartina alterniflora 0.1428 0.0920 0.3879 0.043 40.2 28.4
Unvegetated 0.1171 0.0803 0.7061 0.001 40.2 60.5
Juncus roemerianus 0.1076 0.1121 0.9985 0.230 19.5 11.0
Distichlis spicata 0.0006 0.0026 1.0000 0.001 0.0 0.1
Agalinus maritima 0.0000 0.0000 1.0000 0.0 0.0
Andropogon sp. 0.0000 0.0000 1.0000 0.0 0.0
T-Transition SIMPER results; PERMANOVA p = 0.001
average sd cumulative p mean_start mean_end
Juncus roemerianus 0.1778 0.1068 0.4508 0.001 60.0 32.3
Unvegetated 0.1599 0.0918 0.8562 0.001 34.3 61.4
Spartina alterniflora 0.0183 0.0467 0.9027 0.243 2.2 1.7
Distichlis spicata 0.0098 0.0325 0.9275 0.005 1.2 0.8
Borrichia frutescens 0.0089 0.0146 0.9501 0.002 0.5 1.6
Spartina spartinae 0.0070 0.0187 0.9678 0.320 0.5 0.9
P-Pools/Pannes SIMPER results; PERMANOVA p = 0.702
None
SIMPER not run
H-High Marsh SIMPER results; PERMANOVA p = 0.024
average sd cumulative p mean_start mean_end
Spartina patens 0.1458 0.0901 0.3895 0.013 50.0 21.7
Unvegetated 0.1003 0.0658 0.6573 0.063 38.2 55.4
Panicum virgatum 0.0292 0.0233 0.7352 0.561 5.0 5.4
Cladium jamaicense 0.0222 0.0187 0.7945 0.348 3.3 5.0
Baccharis halimifolia 0.0181 0.0251 0.8427 0.189 2.5 1.7
Agalinus maritima 0.0083 0.0094 0.8650 0.001 0.0 1.7
UE-Upland Edge SIMPER results; PERMANOVA p = 0.2
average sd cumulative p mean_start mean_end
Cladium jamaicense 0.2139 0.1341 0.3571 0.1014 51.7 16.7
Unvegetated 0.1742 0.0773 0.6480 0.0014 26.0 60.8
Spartina patens 0.0708 0.0882 0.7662 0.4014 13.3 2.5
Ilex vomitoria 0.0333 0.0500 0.8219 0.0014 0.0 6.7
Spartina spartinae 0.0264 0.0339 0.8660 0.4014 5.0 0.8
Verbena sp. 0.0167 0.0250 0.8938 0.0014 3.3 0.0
FT-Freshwater Tidal SIMPER results; PERMANOVA p = 0.001
average sd cumulative p mean_start mean_end
Cladium jamaicense 0.1386 0.1062 0.3423 0.002 35.5 16.7
Unvegetated 0.1211 0.0808 0.6414 0.004 64.0 54.3
Spartina patens 0.0600 0.0526 0.7896 0.001 0.0 12.0
Dichanthelium sp. 0.0367 0.0311 0.8801 0.001 0.0 7.3
Panicum virgatum 0.0129 0.0299 0.9120 0.001 0.0 2.6
Juncus roemerianus 0.0083 0.0273 0.9326 0.001 0.0 1.7

NMDS - start/end

In this section, NMDS is performed on data from only the starting and ending years for each vegetation zone. This tab essentially illustrates the PERMANOVA results. For NMDS with all years, see the tab ‘NMDS - all years’.

Non-metric multidimensional scaling is an ordination method that preserves ranked dissimilarities between observations. Exact calculated distances are not preserved in this type of ordination. Points that are closer together on the graphs are more similar than points that are further away, so NMDS is good for seeing groupings and gradients when present. For more information, see the sources referenced below.

This NMDS used Bray-Curtis dissimilarity on the full species matrix (see ‘About’ tab for list of species included), and 3 dimensions.


Final 3-dimensional NMDS stress was 0.0647.

Rules of thumb for interpreting stress, based on the sources below, are:

For more information on NMDS:

Clarke, K. R. (1993). Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology, 18(1), Article 1. https://doi.org/10.1111/j.1442-9993.1993.tb00438.x

Clarke, K. R., & Warwick, R. M. (2001). Change in Marine Communities: An Approach to Statistical Analysis and Interpretation, 2nd ed. - Chapter 5 focuses on NMDS.

Zuur, A. F., Ieno, E. N., & Smith, G. M. (2007). Analysing ecological data. Springer. - Chapter 15 for NMDS.

2-dimensional NMDS plot

(first two axes only)

  • Each small point represents a single vegetation plot at a single time point (start, open circles; or end, filled circles), as in the 3-d plot. Point color represents the plot’s vegetation zone.
  • Large points represent the centroid for each Vegetation Zone/Time period combination. They are labelled with the Vegetation Zone abbreviation and ‘Start’ or ‘End’. Additionally, they are colored by Vegetation Zone and shaped by ‘Start’ vs. ‘End’ time periods. Upon hover, the full Vegetation Zone name and year(s) represented will be provided.
  • The black lines and labels correspond to the red lines and labels in the 3d graph, and represent the loadings: coordinates of species or species groups (centroid of comprising species) specified in the Analysis_Specs worksheet of the veg-specs.xlsx file. Hovering over an arrow will make the species or group name appear more clearly.
  • Some identified loading factors may not appear as arrows; this is usually because all % cover values for that species or group were 0. Cover values may also have been too low to have produced species scores in the NMDS.

Contour plots

The below plots, rather than using arrows for individual species or vegetation groups, create contours for the specific values of each. Contours are labeled with % cover values and are fit as a spline-based surface using vegan::ordisurf().

Click to expand contour plots.

3-d NMDS plot

This 3-d plot is interactive - you can zoom in and rotate the view. Each point represents a single vegetation plot at a single time point (start, open circles; or end, filled circles). Point color represents the plot’s vegetation zone. The red lines and labels represent the coordinates of species or species groups (centroid of comprising species) specified in the Analysis_Specs worksheet of the veg-specs.xlsx file.

Optional additional loadings

No additional loadings specified. If you would like to graph additional environmental factors on the NMDS plot, please specify them in the ‘NMDS additional loadings’ section of the ‘More_Options’ sheet in the veg-specs.xlsx file.

NMDS - all years

By default, this analysis is not run because it may take more computing power than is available to perform ordination on many years worth of data points.

The all-years NMDS was attempted. If the results are not below, it simply didn’t work with your computer.

All measurements at all veg plots (with 3+ years of data) are represented. Centroids are calculated for each zone x year combination. Plots are zoomable. Two plots are provided: the first uses fixed axis scales, to emphasize where zone centroids are relative to each other in ordination space. The second uses free axis scales, to emphasize within-zone differences between years. There may not be much difference between these visually, depending on the spread of points within each zone. Zooming works differently in each. The loadings plot below applies to both graphs (the entire NMDS).

Documentation

R Session Info; click to expand
## R version 4.4.1 (2024-06-14 ucrt)
## Platform: x86_64-w64-mingw32/x64
## Running under: Windows 11 x64 (build 22631)
## 
## Matrix products: default
## 
## 
## locale:
## [1] LC_COLLATE=English_United States.utf8 
## [2] LC_CTYPE=English_United States.utf8   
## [3] LC_MONETARY=English_United States.utf8
## [4] LC_NUMERIC=C                          
## [5] LC_TIME=English_United States.utf8    
## 
## time zone: America/Chicago
## tzcode source: internal
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## other attached packages:
##  [1] scico_1.5.0         metR_0.16.0         patchwork_1.3.0    
##  [4] gt_0.11.1           plotly_4.10.4       ggiraph_0.8.10     
##  [7] ggrepel_0.9.6       vegan_2.6-8         lattice_0.22-6     
## [10] permute_0.9-7       khroma_1.14.0       viridis_0.6.5      
## [13] viridisLite_0.4.2   see_0.9.0           modelbased_0.8.8   
## [16] ggplot2_3.5.1       DT_0.33             kableExtra_1.4.0   
## [19] stringr_1.5.1       readxl_1.4.3        tidyr_1.3.1        
## [22] dplyr_1.1.4         lmerTest_3.1-3      lme4_1.1-35.5      
## [25] Matrix_1.7-0        multcompView_0.1-10 multcomp_1.4-26    
## [28] TH.data_1.1-2       MASS_7.3-60.2       survival_3.6-4     
## [31] mvtnorm_1.3-1       emmeans_1.10.4     
## 
## loaded via a namespace (and not attached):
##  [1] gridExtra_2.3       sandwich_3.1-1      rlang_1.1.4        
##  [4] magrittr_2.0.3      snakecase_0.11.1    compiler_4.4.1     
##  [7] mgcv_1.9-1          systemfonts_1.1.0   vctrs_0.6.5        
## [10] pkgconfig_2.0.3     fastmap_1.2.0       backports_1.5.0    
## [13] labeling_0.4.3      utf8_1.2.4          rmarkdown_2.28     
## [16] nloptr_2.1.1        purrr_1.0.2         xfun_0.48          
## [19] cachem_1.1.0        jsonlite_1.8.9      highr_0.11         
## [22] uuid_1.2-1          broom_1.0.7         parallel_4.4.1     
## [25] cluster_2.1.6       R6_2.5.1            bslib_0.8.0        
## [28] stringi_1.8.4       boot_1.3-30         lubridate_1.9.3    
## [31] jquerylib_0.1.4     cellranger_1.1.0    numDeriv_2016.8-1.1
## [34] estimability_1.5.1  Rcpp_1.0.13         knitr_1.48         
## [37] zoo_1.8-12          parameters_0.24.0   splines_4.4.1      
## [40] timechange_0.3.0    tidyselect_1.2.1    rstudioapi_0.16.0  
## [43] yaml_2.3.10         MuMIn_1.48.4        codetools_0.2-20   
## [46] plyr_1.8.9          tibble_3.2.1        withr_3.0.2        
## [49] bayestestR_0.15.0   coda_0.19-4.1       evaluate_1.0.0     
## [52] isoband_0.2.7       xml2_1.3.6          pillar_1.9.0       
## [55] checkmate_2.3.2     stats4_4.4.1        insight_1.0.0      
## [58] generics_0.1.3      rprojroot_2.0.4     munsell_0.5.1      
## [61] scales_1.3.0        minqa_1.2.8         xtable_1.8-4       
## [64] janitor_2.2.0       glue_1.8.0          lazyeval_0.2.2     
## [67] tools_4.4.1         data.table_1.16.4   forcats_1.0.0      
## [70] grid_4.4.1          crosstalk_1.2.1     datawizard_0.13.0  
## [73] colorspace_2.1-1    nlme_3.1-164        cli_3.6.3          
## [76] fansi_1.0.6         svglite_2.1.3       gtable_0.3.5       
## [79] sass_0.4.9          digest_0.6.37       pbkrtest_0.5.3     
## [82] htmlwidgets_1.6.4   farver_2.1.2        memoise_2.0.1      
## [85] htmltools_0.5.8.1   lifecycle_1.0.4     httr_1.4.7         
## [88] here_1.0.1